RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
Low Power CMOS Analog Multipliers
Supervisor:
Dr. C. Chen
Student:
Zheng Li
Department of Electrical and Computer Engineering University of Windsor
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
1. Application of Analog Multipliers Analog multiplier is an important subcircuit for many applications such as adaptive filters, frequency doublers, and modulators. It performs linear product of two continuous signals x and y, yielding an output z = Kxy, where K is a constant with suitable dimension[1]. Z=Kxy
Z=Kxy
Signal x
Signal x
Signal y
Signal y
Fig 1 The Application of Analog Multiplier
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
2. Performance Metrics of Analog Multipliers The linearity, supply voltage, power dissipation and noise are the main metrics of performance [3]. We try to design some specific structures or topologies for the analog multiplier that have low power dissipation while at the same time keeping good linearity, low supply voltage and low noise. Noise
Linearity
Power dissipation Input/output impedance Speed
Gain Supply Voltage Voltage swings
Fig 2 The Metrics of Analog Multiplier
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
3. Basic Idea of Our Low Power Design Use total current to represents the power consumption x 1 I D K(VGS VTH )2 (1 VDS ) 2 1 I D K[(VGS VTH )VDS VDS2 ] 2
[3]Where K = oCox W/L and
x
y
y Fig. 3. Power consumption and input range for (a) series structure and (b) parallel structure
VTH are the transconductance parameter and the threshold voltage of the device, respectively, and represents the channel-length modulation effect for long channel devices. By biasing the transistors to operate in the triode region, one can reduce the drain current while keeping a relatively large input range.
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
4. Design Flow of Our structures Theoretically Computation and Analysis Simulation in the Schematic Level and Verification Layout Layout Verification ( DRC, LVS and Extraction) Simulation in the Layout Level and Verification Fabrication and Testing
(not finished)
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
5. Our First Multiplier structure VDD=1.5V Vbias=1V
Vo1
i1
i2
i3
i4
X+x
Vo2 X+x
X-x Y+y
Y-y
Fig. 4. First Multiplier Structure
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
6. Theoretical Analysis of First structure For the transistors working in the triode region, we have the equation for small signal model as id Kvgs vDS , therefore, we have: i1 KvDS1 ( x y )
vDS 1 vo1 Y y
i2 KvDS 2 (x y)
vDS 2 vo 2 Y y
i3 KvDS 3 ( x y )
vDS3 vo1 Y y
i4 KvDS4(x y)
vDS 4 vo 2 Y y
By this equations, we have (i1 i3 ) (i2 i4 ) 4Kxy The bias conditions is
vo X x VTH 0 Y y X x VTH
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
7. Linearity Analysis of First structure
Fig. 5.(a) DC Response for signal x (with body effect)
Fig. 5.(b) DC Response for signal y (with body effect)
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
7. Linearity Analysis of First structure Linearity Error [%]
1 0.5 0 0.2
0.3
0.4
2x [V]
Fig. 6.(a) Linearity Error of signal x with 2y=0.4V (with and without body effect) Linearity Error [%]
3 2 1 0 0.2
0.3
0.4
2y [V]
Fig. 6.(b) Linearity Error of signal y with 2x=0.4V (with and without body effect)
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
8. Power Consumption of First Structure Power consumption can be estimated by the total supply current if the supply voltage is constant. Because we have Itotal i1 i2 i3 i4 1 i1 i3 2 Kl ( X Y VTH )(Vo1 Y ) 2 Kl ( x y ) y Ku (Vb Vo1 VTH )2 2 1 i2 i4 2 Kl ( X Y VTH )(Vo 2 Y ) 2 Kl ( x y) y Ku (Vb Vo 2 VTH ) 2 2 2 Therefore Itotal 2 Kl ( X Y VTH )(Vo1 Vo 2 2Y ) 4 Kl y
We could decrease the power consumption by decreasing the (X-Y), K l and signal y.
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
8. Power Consumption of First Structure I total
I total 40u
30u
30u 20u 20u 10u 0
10u 0.7
0.9
1.1
1.3
X-Y [V]
Fig. 7.(a) Total Current Versus X-Y
Fig. 7.(c) Total Currents for different signal distribution
0
1
2
3
Kl
4
Fig. 7.(b) Total Current VersusK
l
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
9. Noise Analysis of First Structure The total output noise of Fig. 5 is given by[4] 1 in2;o 4in2;lin 2in2; sat 16 KT ( g ds.l g m ,u )df 3
gds.l Kl ( X Vo VTH )
and
where
gm,u 2 Kl Ku ( X Y VTH )(Vo Y )
[2] Hence the input noise is vn2;i
in2;o 16 Kl2
kT 2 Ku [( X Vo VTH ) ( X Y VTH )(Vo Y )] Kl 3 Kl
That implies that Ku/K l and (X-Y) should be decreased to improve the noise performance.
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
9. Noise Analysis of First Structure input-noise [aV2/Hz]
0.25
0.33
0.5
1
Fig 8(a) The input-referred noise voltage of Fig. 2
Ku/K l versus Ku/K l
input-noise [aV2/Hz]
0.7
0.9
1.1
1.3
1.5
(X-Y) V
Fig 8(b) The input-referred noise voltage of Fig. 2 versus X-Y
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
10. Layout of First Multiplier structure
Fig. 9 Layout of our first multiplier structure
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
10. Layout of First Multiplier structure When we do the layout of the structures, we consider to: •
Decrease the distance among the transistors as long as we could pass the DRC.
•
Try to decrease the length of the metal or polysilicon to decrease the parasite resistors and capacitors involved.
•
For the differential structure, try to make involved parasite resistors and capacitors symmetric.
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
11. Extracted Structure
Fig. 10 Extracted First Multiplier Structure
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
12. Comparison before and after layout
Fig. 11 (a) The comparison of total current before and after layout THD[%]
100K
Fig. 11 (b) The comparison of AC response before and after layout THD[%]
1M
100M
1G
Frequency
Fig. 11 (c) The comparison of THD for signal 2x=0.4V before and after layout with frequency
100K
1M
100M
1G
Frequency
Fig. 11 (d) The comparison of THD for signal 2y=0.4V before and after layout with frequency
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
13. Our Second Multiplier structure VDD=1.5V
X+x
iP1
iP 2 Vo1
iP 3
iP 4
X-x
Vo2
Y+y
Y-y
Ground Fig. 12. Second Multiplier Structure
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
14. Theoretical Analysis of Second structure For the PMOS transistors, we have: 1 iP1 KP (VDD X x VTHP )2[1 (VDD VP1)] 2
iP 2
1 K P (VDD X x VTHP )2 [1 (VDD VP 2 )] 2
1 iP3 KP (VDD X x VTHP )2 [1 (VDD VP3 )] 2
iP 4
1 K P (VDD X x VTHP )2 [1 (VDD VP 4 )] 2
For the transistors N1-N4, we have: 1 i1 KN[(Y y VP1 VTHN )(VO1 VP1) (VO1 VP1)2 ] 2
1 i2 KN [(Y y VO1 VTHN)(VP2 VO1 ) (VP2 VO1 )2 ] 2
1 i3 KN[(Y y VP3 VTHN )(VO2 VP3) (VO2 VP3)2 ] 2
1 i4 KN [(Y y VO2 VTHN)(VP4 VO2 ) (VP4 VO2 )2 ] 2
And also, we realized that: iP1 i1 iM 1
iP 2 i2 iM 2
iP 3 i3 iM 3
iP 4 i4 iM 4
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
14. Theoretical Analysis of Second structure The transistors M1-M4 can be considered as the resistors with resistance approximately equal to R K (V 1 V ) ON
M
Solving the equations above with i1 i2 gives the following approximate result: V O 1 VO 2
DD
THN
i3 i4
KN KM xy KP
which is a multiplication of two input signals, x and y The bias conditions is VP VTHP X x VDD VTHP
For P1-P4
Y y VO VTHN
For N1-N4
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
15. Linearity Analysis of Second structure
Fig. 13.(a) DC Response for signal x (with body effect)
Fig. 13.(b) DC Response for signal y (with body effect)
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
15. Linearity Analysis of Second structure Linearity Error [%]
1.5 1 0.5 0
0.2
0.3
0.4
2x [V]
Fig. 14.(a) Linearity Error of signal x with 2y=0.4V (with and without body effect) Linearity Error [%]
0.6 0.4 0.2 0.1 0
0.2
0.3
0.4
2y [V]
Fig. 14.(b) Linearity Error of signal y with 2x=0.4V (with and without body effect)
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
16. Power Consumption of Second Structure Power consumption can be estimated by total current: Itotal iP1 iP 2 iP3 iP 4
Therefore
1 1 Itotal KP(VDD X xVTHP )2[1(VDD VP1)] KP(VDD X xVTHP )2[1(VDD VP2)] 2 2 1 1 2 KP(VDD X xVTHP ) [1(VDD VP3)] KP(VDD X xVTHP )2[1(VDD VP4)] 2 2
Approximately, we have I total 2 K P [(VDD X VTHP ) 2 x 2 ] We see that the power dissipation has nothing to do with the signal y and DC bias Y for transistors N1-N4.
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
16. Power Consumption of Second Structure I total [uA]
I total [uA] when a=W/L=0.35/0.7u 80
30
60
20
40
10
20
0
0 0.4
0.45
0.5
0.55
0.6
X [V]
Fig. 15.(a) Total Current Versus X
I total[uA]
a
2a
3a
4a
Size of
KP
Fig. 15.(b) Total Current VersusK P
I total [uA] when X=0.5V and Y=1.5V, 2y=0.4V
20
20 10
0
10
0 1.5
1.45
1.4
1.35
Y [V]
Fig. 15.(c) Total Current Versus Y
0
0.1
0.2
0.3
0.4
Fig. 15.(d) Total Current Versus 2y
2y [V]
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
17. Noise Analysis of Second Structure The total output noise of Fig. 11 is given by i 8i 4i 2 n;o
2 n;lin
2 n;sat
32 16kTgds1df 16kTgds2df kTgmdf 3
where
gds1 KN (Y VO VTHN )
gds2 KM(VDD VP VTHN) gm KP (VDD X VTHP )
Then, the total input noise would be 2 n ;o
v
8v
2 n ;lin
4v
2 n ;sat
K P2 K P2 32 K P4 16kT 2 df 16kT 2 df kT 2 2 df 3 KM KN KM KN
This suggests that KM , KN should be increased and KP should be decreased to improve the noise performance
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
17. Noise Analysis of Second Structure input-noise [aV2/Hz] KP
180 160 140 120 100 80 60
KM
40 20 0
KN 1
2
3
4
3 1.05
4 1.10
5
Fig 16(a) The input-referred noise voltage of Fig. 2 versus KP, KM and KN input-noise 1 1 1 1 1 1 1 1 1 1
2 2 2 2 2 2 2 2 2 1
[aV2/Hz] 8 7 6 5 4 3 2 1 0 9 1 0.95
2 1.00
5
YX (V)
Fig 16(b) The input-referred noise voltage of Fig. 2 versus Y-X
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
18. Layout of Second Multiplier structure
Fig. 17 Layout of our second multiplier structure
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
19. Extracted Structure
Fig. 18 Extracted Second Multiplier Structure
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
20. Comparison before and after layout
Fig. 19 (a) The comparison of total current before and after layout THD[%]
100K
Fig. 19 (b) The comparison of AC response before and after layout THD[%]
1M
100M
1G
Frequency
100K
1M
100M
1G
Frequency
Fig. 19 (c) The comparison of THD for signal Fig. 19 (d) The comparison of THD for signal 2x=0.4V before and after layout with frequency 2y=0.4V before and after layout with frequency
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
21. Most recommended structure by [3]
Vbias=1V
VDD=1.5V
Vo1
Vo2
X+x
X-x Y-y
Y+y Ground Fig. 20 The most recommended multiplier structure by[2]
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
22. Comparison of Three Structures Our First Structure
Our Second Structure
Most Recommended by[2]
Power Consumption
22.65u
27.15u
68.31u
Estimated
12u*12u
27u*16u
0.8%
1.5%
3.3%
2.6%
0.48%
0.9%
0.43%
0.4%
0.4%
Area Linearity Error of x Linearity Error of y THD of x
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
22. Comparison of Three Structures Our First Structure
Our Second Structure
Most Recommended by[2]
THD of y
0.51%
0.8%
1.1%
Noise
76a
128a
109a
Band width
2G
80M
80M
Number of Transistors
6
12
10
Technology
CMOS 0.35
CMOS 0.35
CMOS 0.35
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
Conclusion •
By Biasing the transistors in the linear region and decrease the drainsource voltage, we could get low current while at the same time, keeping a relatively wider input range. This is the basic idea of our low power design.
•
We try to use transistors as few as possible if we could satisfy the other performance metrics in order to get less area fabrication.
•
Body effect is an important second-order effect which we have to consider for good linearity.
•
In the layout level, for the differential structure, try to make involved parasite resistors and capacitors symmetric.
•
Post-layout simulation is important for performance verification.
•
Our two structures have less power consumption compared with the most recommended structure in [2] while keeping other performance good.
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
Reference [1] K. Bult and H. Wallinga, “A four-quadrant analog multiplier,” IEEE Journal of Solid-State Circuits, vol. Sc-21, no.3, pp. 430-435, June 1986. [2] G. Han and E. Sanchez-Sinencio, “CMOS transconductance multipliers: A tutorial,” IEEE Transactions on Circuits and Systems-II: Analog and Digital Signal Processing, vol. 45, no. 12, pp. 1550-1563, December 1998. [3] B. Razavi, “Design of analog CMOS integrated circuits,” New York: McGraw-Hill, 2001. [4] K.R. Laker and W. M. C. Sansen, “Design of Analog Integrated Circuits and Systems,” New York: McGraw-Hill, 1994.
And another about 70 papers in the Literature survey. The papers I have submitted to the conferences: • A Low Power and Lower Noise CMOS Analog Multiplier • A Low Power CMOS Analog Multiplier
RESEARCH CENTRE FOR INTEGRATED MICROSYSTEMS – UNIVERSITY OF WINDSOR
Thanks a lot!